decision-capture
GitHub自动检测对话中的决策,记录理由、备选方案及结果跟踪,持久化存储以形成可追溯的知识资产。
Trigger Scenarios
Install
npx skills add nearai/ironclaw --skill decision-capture -g -y
SKILL.md
Frontmatter
{
"name": "decision-capture",
"version": "0.2.0",
"activation": {
"tags": [
"commitments",
"decision-making"
],
"keywords": [
"decided",
"decision",
"chose",
"going with",
"settled on",
"picked",
"landed on",
"went with",
"finalized",
"agreed on",
"opted for",
"concluded",
"confirmed",
"final answer",
"made the call",
"ruling",
"verdict",
"record decision"
],
"patterns": [
"(?i)(we|I|team) (decided|chose|went with|picked|settled on|landed on|opted for)",
"(?i)let's go with",
"(?i)the (decision|call|verdict) is",
"(?i)record (this|that) decision",
"(?i)(instead of|replaced|switched to|migrated to)"
],
"exclude_keywords": [
"undecided",
"considering",
"thinking about",
"tentative",
"maybe"
],
"max_context_tokens": 1200
},
"description": "Detect decisions in conversation and record them with rationale, alternatives, and outcome tracking."
}
Decision Capture
When a decision is detected in conversation, record it in the commitments workspace. Decisions are durable knowledge — they explain why a path was chosen and enable outcome tracking over time.
Detection
Look for explicit decision language:
- "We decided to..." / "I'm going with..." / "Let's do X"
- "The call is..." / "We settled on..."
- "I chose X over Y because..."
Do NOT capture:
- Brainstorming or hypotheticals: "maybe we should...", "what if we..."
- Preferences without commitment: "I think X is better"
- Questions: "should we go with X?"
When uncertain, ask: "Was that a decision, or still thinking it through?"
Recording
This skill is only successful if the decision is actually persisted. Do not just summarize or acknowledge the decision.
Execution order is mandatory:
- Call
memory_writeforprojects/commitments/decisions/<date>-<slug>.md - If applicable, call
memory_writefor a follow-on commitment inprojects/commitments/open/ - If applicable, call
memory_writeforcontext/intel/<slug>.md - Only then confirm to the user what was recorded
Never say a decision was "captured", "recorded", or "saved" unless the
corresponding memory_write call succeeded.
Write to projects/commitments/decisions/<date>-<slug>.md via memory_write:
---
type: decision
decided_at: <today YYYY-MM-DD>
context: <topic-slug>
participants: [<who was involved>]
confidence: <high if explicit, medium if inferred>
reversible: <true|false>
outcome: null
outcome_positive: null
tags: [<relevant tags>]
---
# <What was decided>
## Context
<Why this decision was needed — 1-2 sentences.>
## Options considered
1. **<Option A>** — <brief pros/cons>
2. **<Option B>** — <brief pros/cons>
## Rationale
<Why this option was chosen.>
## Outcome
<To be filled in later when outcome is known.>
Follow-through
- If the decision creates an obligation (e.g., "we decided to migrate by Q2"), also create a commitment in
projects/commitments/open/following the commitment schema. - Write an intelligence MemoryDoc to
context/intel/<slug>.mdwith a brief summary: "Decided X on. Rationale: ." This makes the decision searchable as durable knowledge.
For explicit requests like "record this decision", "log this decision", or "note the decision", default to doing all required writes immediately rather than asking a follow-up question unless the content is genuinely ambiguous.
Outcome tracking
The triage mission checks for decisions older than 7 days without an outcome. It prompts: "You decided <X> <N> days ago. How did it turn out?" When the user provides an outcome, update the decision file's outcome and outcome_positive fields.
Confirmation
After the write(s) succeed, briefly confirm:
- where the decision was written
- whether any follow-on commitment was created
- one-line summary of the rationale
Version History
- 5380a32 Current 2026-08-20 07:09


